The browser version you are using is not recommended for this site.
Please consider upgrading to the latest version of your browser by clicking one of the following links.

We are sorry, This PDF is available in download format only

Universityresearchteam buildsamulti-purposeHPC clustertoanalysedatafrom CERN’sATLASproject

• HPC “Thisresearchis crucialifwe’reto understandhowthe universeworks.If theconstituentsof darkmatterarenew particles,ATLASshould discoverthemandour workherecouldhelp explainthemysteryof darkmatter.” DrFabrizioSalvatore,Experimental ParticlePhysics,UniversityofSussex Customerprofile Company: UniversityofSussex Industry: HigherEducation Country: UnitedKingdom Employees: 2160 Website: www.sussex.ac.uk Businessneed TheUniversityofSussexneededapowerfulhighperformance computing(HPC)clustertoprocessdatacomingoutofCERN’sATLAS projectandofferaffordablecomputefacilitiesforalldepartments. Solution TheuniversityworkedwithDelltodesignanddeployaHPCcluster withaparallelfilesystem,deliveringfastaccesstodataandeasy managementforalldepartments. Benefits • HPCclusterisnowamodelfor otheruniversitieslookingathigh performancefilesystems • Workthatpreviouslytook20hours iscompletedinfiveminutes • Clusterisusedtoconductcrucial researchforCERN’sATLASproject • Otherdepartmentshavecost- effectiveaccesstoHPC • Strongprojectmanagementensures deploymentin11workingdays TheworkconductedattheEuropean OrganizationforNuclearResearch(CERN) issomeofthemostimportantscientific researchundertakentoday.Witharemitto findevidenceofnewphysics–behaviour beyondwhatiscurrentlyunderstoodin thestandardmodelofphysics–CERN issearchingfordarkmatterandsuper- symmetrybetweensub-atomicparticles. “Nowthere’snoneed tostrugglewithday- to-dayanalytics.Asan example,oneofmy studentscametosee meandexplainedthat he’djustcompleteda taskinfiveminutes– previously,thesame typeoftaskwouldhave takenhim20hours.” DrFabrizioSalvatore,Experimental ParticlePhysics,University ofSussex To do this, the ATLAS project measures the results of proton interactions after high-speed collisions. Each time the Large Hadron Collider (LHC) fires protons at each other, a huge quantity of data is collected. This data needs to be analysed and interpreted by teams at universities around the world, including the University of Sussex. Dr Fabrizio Salvatore, from the Experimental Particle Physics team at the University of Sussex, explains: “We’re a tier-three research facility looking at data produced by CERN – this means we do some of the most in-depth analysis and need huge computational power to complete our work. By analysing this data, we hope to learn about the basic forces that have shaped our universe since the beginning of time – and will determine its fate.” Whentheteamfirststartedworking ontheATLASproject,therewasno suitablehighperformancecomputing (HPC)clusterattheUniversityof Sussex,soFabrizioSalvatoreand JeremyMaris,ResearchComputing Managerattheuniversity,started thinkingaboutasolution.Initially,they consideredbuildingaHPCclusterfor usebyjusttheExperimentalParticle Physicsteam.Buttheysoonrealised thatwithadditionalmoneyfromthe universitytheycouldbuildalarger HPCclusterthatotherteamswould beabletousetoo.Marisexplains:“We neededtobuildaclusterthatcould dotwothings.First,enablethephysics departmenttoanalysedatacoming fromtheATLASproject.And,second, offercomputefacilitiestoother researchteamssuchaschemistry, maths,economics,andlifesciencesin otheruniversitydepartments.” Choosinganexperienced solutionpartner In practical terms, this meant a fast HPC cluster, with lots of storage capacity and fast data access. Maris decided that a parallel file system was the best option to handle the quantities of data moving through the cluster, and started talking to different IT partners about building the cluster. Maris had worked with Dell before, and the university had used Dell desktops and laptops for some time, but never at a HPC level. As part of a pre-tender exercise, Maris and his team spoke to Dell about the Technology at work Services DellConsultingServices DellProSupport −NextBusinessDayOnsite Service Hardware ™ ™ Dell PowerEdge R410/R510/ ® R610/R710serverswithIntel ® Xeon processors5650 DellPowerEdgeR815serverswith ® ® AMD Opteron 6172processors ™ MD1200and DellPowerVault MD3200directattachedstorage ™ QLogic Infiniband Software ™ InfinibandFabricSuite ® BrightClusterManager ™ Lustre filesystem challenges ahead, and what a suitable HPC cluster might look like. Maris says: “I had a good idea of what we needed, but it was great to get Dell’s input right at the start. Dell’s on- going involvement at CERN – and particularly the ATLAS project – was really useful at this stage.” After a tender exercise conducted through the National Storage and Servers Agreement (NSSA), the university chose to work with the Dell Consulting team to create the new HPC cluster. Maris says: “We had a number of meetings with Dell. The technical consultants were impressive and understood our needs.” PowerfulIntelprocessorsdeliver evenhigherperformance TheclusterthatMarisandSalvatore ™ ™ createdisbasedonDell PowerEdge ® ® R410serverswithIntel Xeon processors5650workingascompute nodesforaPowerEdgeR610server thatoperatesasaheadnode.Two PowerEdgeR510soperateashome fileservers,andadditionalnodes usePowerEdgeR815serverswith ® ® AMD Opteron 6172processors.The ™ Lustre filesystemusesPowerEdge ™ R510scoupledwithDellPowerVault MD1200s.Thehighlyavailable metadataserverusestwoPowerEdge R710sandDellPowerVaultMD3200 directattachedstorage.Allnodes ™ areinterconnectedwithQLogic Infinibandcards.Andeveryelementis protectedwithDellProSupportNext BusinessDayOnsiteService. Marissays:“AtCERN,theyuseIntel processorsinthemajorityoftheir computenodes.ThePowerEdgeR410 serversrunningIntelXeonprocessors 5650offerexcellentperformance, whiletheR815sprovideacost- effectivesolutionformultithreaded applicationsthatrequirealotof Read the full Universityresearchteam buildsamulti-purposeHPC clustertoanalysedatafrom CERN’sATLASproject.

Related Videos